Insights on AI adoption
What I've learned helping teams use AI well
Notes from the programmes I run, the teams I lead and the participants who told me what changed. Start with the first article, then pick the topic in front of you.
AI adoption after the licences: what leadership teams actually have to change
AI adoption stalls when nothing changes after the licences are bought. Here is what leaders must change so engineering teams practise on their real work.
-
Why AI training works better with the whole team
This guide to AI training for engineering managers explains why I train developers, testers, analysts, designers and product people together on one shared spec.
-
Can a non-developer use Claude Code?
Claude Code for non-technical people starts with real work. See the tasks marketers, PMO, QA and admin staff brought to my training and where they got stuck.
-
Context matters: why more instructions can make AI output worse
More instructions can make AI output worse. These CLAUDE.md best practices keep the file small, decide where knowledge lives and prove Claude is reading it.
-
How to check AI output before you trust it
This is how to verify AI output without rereading every line. Keep review apart from verification, ask what proves the work, and automate the checks you repeat.
-
Turn repeated work into a reusable skill
Claude skills for business turn repeated jobs into written procedures. Find the work you repeat, build from your own transcripts and judge any skill you borrow.
-
What can you safely let an AI agent run without you?
AI agents for business owners can save time on the right jobs. Learn three tests a job must pass and what to know before an agent reads your email and files.
-
How to measure AI adoption in a software team
Token counts and commit totals prove nothing about learning. This is how to measure AI adoption in software teams with situation questions and a baseline.
-
Coaching AI adoption from how your team actually uses it
Surveys show what people believe they do. Claude Code usage analytics show what they actually do. I capture sessions with consent and coach the real gaps.